SPLH 861 Example 8 page 1 Example 8: Path Analysis for Mediation (complete syntax and output available for SAS, Mplus, and STATA electronically) Internal Motivation to Respond without Sexism (M1) Figure 1 from: Gervais, S. J. & Hoffman, L. (2013). Just think about it: Mindfulness, sexism, and prejudice towards feminists. Sex Roles, 68(5), 283-295. Mindfulness (X) External Motivation to Respond without Sexism (M2) Warmth Towards Feminists (Y) A sample of 653 undergraduates completed the six measures depicted in Figure 1 (residual covariances among the mediators are not shown for diagram clarity). Table 3 shows the correlations of the six variables by gender. Hostile Sexism (M3) Benevolent Sexism (M4) The research questions were as follows: (1) To what extent do these four mediators account for the relationship between mindfulness and warmth towards feminists? (2) How do these direct and indirect effects differ by gender? Accordingly, we will begin with a single-group model, and then examine a multiple-group model in which all parameters are estimated separately for men and women. From there, one would proceed by constraining specific direct and indirect effects to be equal across genders and note the decrease in model fit in doing so. SAS PROC CALIS Single-Group Path Model: TITLE1 "Single-Group Path Model"; PROC CALIS DATA=work.Mindfull MEANSTR METHOD=FIML; * VAR = List of variables in model; VAR MindC Intern Extern Hostile Benev Nontrad; * MEAN = Labeling means/intercepts per variable; MEAN MindC=Xint, Intern=M1int, Extern=M2int, Hostile=M3int, Benev=M4int, NonTrad=Yint; * PVAR = Labeling variances/residual variances per variable; PVAR MindC=Xvar, Intern=M1var, Extern=M2var, Hostile=M3var, Benev=M4var, NonTrad=Yvar; * PCOV = Requesting and labeling residual covariances; PCOV Intern Extern=CovM12, Intern Hostile=CovM13, Intern Benev=CovM14, Extern Hostile=CovM23, Extern Benev=CovM24, Hostile Benev=Cov34;
* PATH = Model specification and labels; PATH MindC ---> NonTrad = mxtoy, MindC ---> Intern Extern Hostile Benev = XtoM1 XtoM2 XtoM3 XtoM4, Intern Extern Hostile Benev ---> NonTrad = M1toY M2toY M3toY M4toY; * TESTFUNC = Requesting indirect effects; TESTFUNC XtoM1toY XtoM2toY XtoM3toY XtoM4toY; XtoM1toY = XtoM1*M1toY; XtoM2toY = XtoM2*M2toY; XtoM3toY = XtoM3*M3toY; XtoM4toY = XtoM4*M4toY; RUN; TITLE1; SPLH 861 Example 8 page 2 MPLUS Single-Group Path Model (see syntax for full code):! Model code: ON = Y ON X, WITH = correlation! Labels in parentheses (can be used to name constraints between groups) MODEL:! Bring X into the likelihood by estimating its mean and variance [Mind1C] (Xint); Mind1C (Xvar);! Intercepts and residual variances for other variables [Intern Extern Hostile Benev NonTrad] (M1int M2int M3int M4int Yint); Intern Extern Hostile Benev NonTrad (M1var M2var M3var M4var Yvar);! Direct Mind1C --> NonTrad NonTrad ON Mind1C (XtoY);! Left side of model Intern ON Mind1C (XtoM1); Extern ON Mind1C (XtoM2); Hostile ON Mind1C (XtoM3); Benev ON Mind1C (XtoM4);! Right side of model NonTrad ON Intern (M1toY); NonTrad ON Extern (M2toY); NonTrad ON Hostile (M3toY); NonTrad ON Benev (M4toY);! Residual covariances among mediator variables Intern Extern Hostile Benev WITH Intern Extern Hostile Benev;! Testing indirect effects MODEL CONSTRAINT: NEW (XtoM1toY XtoM2toY XtoM3toY XtoM4toY); XtoM1toY = XtoM1*M1toY; XtoM2toY = XtoM2*M2toY; XtoM3toY = XtoM3*M3toY; XtoM4toY = XtoM4*M4toY; STATA SEM Single-Group Path Model: * /// means continue the command + comment * // means comment only * sem, standardized // print standardized effects this time * first ran using coeflegend, which displays parameter labels (see log) display as result "Single group path model omitting coeflegend" display as result "Adding indirect effects (that use parm labels)" sem (mindc -> nontrad) /// XtoY (mindc -> intern extern hostile benev ) /// XtoM1,M2,M3,M4 (intern extern hostile benev -> nontrad), /// M1,M2,M3,M4toY means(mindc) /// print mean var(mindc) /// print variance covariance(e.intern*e.extern e.intern*e.hostile) /// residual covariances covariance(e.intern*e.benev e.extern*e.hostile) /// residual covariances covariance(e.extern*e.benev e.hostile*e.benev) /// residual covariances method(mlmv) // full-information ML estat eqgof, // R2 per variable estat gof, stats(all), // model fit nlcom _b[intern:mindc]*_b[nontrad:intern] // indirect effect XtoM1toY nlcom _b[extern:mindc]*_b[nontrad:extern] // indirect effect XtoM2toY nlcom _b[hostile:mindc]*_b[nontrad:hostile] // indirect effect XtoM3toY nlcom _b[benev:mindc]*_b[nontrad:benev] // indirect effect XtoM4toY
STATA SEM unstandardized output for the Single-Group Path Model: Structural equation model Number of obs = 653 Estimation method = mlmv Log likelihood = -5410.7727 ---------- OIM ---------- Structural nontrad <- intern.5626147.0731554 7.69 0.000.4192328.7059967 extern.0578879.0760251 0.76 0.446 -.0911186.2068944 hostile -.8125585.107334-7.57 0.000-1.022929 -.6021878 benev -.2124246.107228-1.98 0.048 -.4225876 -.0022615 -.0119824.1889878-0.06 0.949 -.3823917.3584269 _cons 7.456182.7308971 10.20 0.000 6.02365 8.888714 -------- intern <-.3353112.1164601 2.88 0.004.1070536.5635687 _cons 4.971255.1099664 45.21 0.000 4.755725 5.186785 -------- extern <-.041326.1079943 0.38 0.702 -.170339.252991 _cons 4.06338.1017661 39.93 0.000 3.863922 4.262838 -------- hostile <- -.1958762.0742917-2.64 0.008 -.3414852 -.0502672 _cons 4.069324.0701493 58.01 0.000 3.931834 4.206814 -------- benev <- -.0521908.0699051-0.75 0.455 -.1892022.0848206 _cons 4.108647.0660072 62.25 0.000 3.979276 4.238019 ---------- mean(mindc).8345001.0172896 48.27 0.000.8006132.8683871 ---------- var(e.nontrad) 4.400699.2446747 3.946351 4.907358 var(e.intern) 1.728817.095677 1.551106 1.926889 var(e.extern) 1.477754.0821223 1.325253 1.647804 var(e.hostile).7035186.0389344.6312015.7841211 var(e.benev).6228914.0344723.5588622.6942564 var(mindc).195201.0108029.1751356.2175653 ---------- cov(e.intern,e.extern).6021503.0673758 8.94 0.000.4700961.7342044 cov(e.intern,e.hostile) -.3739513.045571-8.21 0.000 -.4632688 -.2846338 cov(e.intern,e.benev) -.0067172.04061-0.17 0.869 -.0863114.072877 cov(e.extern,e.hostile).0364342.0401216 0.91 0.364 -.0422026.1150711 cov(e.extern,e.benev).146955.0380589 3.86 0.000.0723609.221549 cov(e.hostile,e.benev).1118392.0262723 4.26 0.000.0603464.163332 ---------- LR test of model vs. saturated: chi2(0) = 0.00, Prob > chi2 =. Equation-level goodness of fit Variance depvars fitted predicted residual R-squared mc mc2 observed nontrad 5.883612 1.482912 4.400699.2520411.502037.2520411 intern 1.750765.0219471 1.728817.0125357.1119632.0125357 extern 1.478088.0003334 1.477754.0002255.0150181.0002255 hostile.711008.0074894.7035186.0105335.1026326.0105335 benev.6234231.0005317.6228914.0008529.0292041.0008529 overall.0176456 mc = correlation between depvar and its prediction mc2 = mc^2 is the Bentler-Raykov squared multiple correlation coefficient SPLH 861 Example 8 page 3
---------------------------------------------------------------------------- Fit statistic Value Description Likelihood ratio chi2_ms(0) 0.000 model vs. saturated p > chi2. chi2_bs(15) 439.601 baseline vs. saturated p > chi2 0.000 Population error RMSEA 0.000 Root mean squared error of approximation 90% CI, lower bound 0.000 upper bound 0.000 pclose 1.000 Probability RMSEA <= 0.05 Information criteria AIC 10871.545 Akaike's information criterion BIC 10983.585 Bayesian information criterion Baseline comparison CFI 1.000 Comparative fit index TLI 1.000 Tucker-Lewis index Size of residuals CD 0.018 Coefficient of determination ---------------------------------------------------------------------------- Note: SRMR is not reported because of missing values. SPLH 861 Example 8 page 4. nlcom _b[intern:mindc]*_b[nontrad:intern] // indirect effect XtoM1toY _nl_1: _b[intern:mindc]*_b[nontrad:intern] _nl_1.188651.0699633 2.70 0.007.0515254.3257765. nlcom _b[extern:mindc]*_b[nontrad:extern] // indirect effect XtoM2toY _nl_1: _b[extern:mindc]*_b[nontrad:extern] _nl_1.0023923.0070151 0.34 0.733 -.0113572.0161417. nlcom _b[hostile:mindc]*_b[nontrad:hostile] // indirect effect XtoM3toY _nl_1: _b[hostile:mindc]*_b[nontrad:hostile] _nl_1.1591609.0639227 2.49 0.013.0338747.2844471. nlcom _b[benev:mindc]*_b[nontrad:benev] // indirect effect XtoM4toY _nl_1: _b[benev:mindc]*_b[nontrad:benev] _nl_1.0110866.0158691 0.70 0.485 -.0200162.0421894
SAS PROC CALIS Multiple-Group Path Model (all parameters separate by gender): SPLH 861 Example 8 page 5 TITLE1 "Multiple-Group Path Model"; PROC CALIS DATA=work.Mindfull MEANSTR METHOD=FIML; * VAR = List of variables in model; VAR MindC Intern Extern Hostile Benev Nontrad; * GROUP creates separate models (referred to by number); GROUP 1 / LABEL="Men" DATA=work.Men; GROUP 2 / LABEL="Women" DATA=work.Women; * Model for Men -- all parameters are now labeled separately for full non-invariance; MODEL 1 / GROUP=1; MEAN MindC=mXint, Intern=mM1int, Extern=mM2int, Hostile=mM3int, Benev=mM4int, NonTrad=mYint; PVAR MindC=mXvar, Intern=mM1var, Extern=mM2var, Hostile=mM3var, Benev=mM4var, NonTrad=mYvar; PCOV Intern Extern=mCovM12, Intern Hostile=mCovM13, Intern Benev=mCovM14, Extern Hostile=mCovM23, Extern Benev=mCovM24, Hostile Benev=mCov34; PATH MindC ---> NonTrad = mxtoy, MindC ---> Intern Extern Hostile Benev = mxtom1 mxtom2 mxtom3 mxtom4, Intern Extern Hostile Benev ---> NonTrad = mm1toy mm2toy mm3toy mm4toy; * Model for Women -- all parameters are now labeled separately for full non-invariance;; MODEL 2 / GROUP=2; MEAN MindC=wXint, Intern=wM1int, Extern=wM2int, Hostile=wM3int, Benev=wM4int, NonTrad=wYint; PVAR MindC=wXvar, Intern=wM1var, Extern=wM2var, Hostile=wM3var, Benev=wM4var, NonTrad=wYvar; PCOV Intern Extern=wCovM12, Intern Hostile=wCovM13, Intern Benev=wCovM14, Extern Hostile=wCovM23, Extern Benev=wCovM24, Hostile Benev=wCov34; PATH MindC ---> NonTrad = wxtoy, MindC ---> Intern Extern Hostile Benev = wxtom1 wxtom2 wxtom3 wxtom4, Intern Extern Hostile Benev ---> NonTrad = wm1toy wm2toy wm3toy wm4toy; * Indirect effects for both groups; TESTFUNC mxtom1toy wxtom1toy mxtom2toy wxtom2toy mxtom3toy wxtom3toy mxtom4toy wxtom4toy; mxtom1toy = mxtom1*mm1toy; wxtom1toy = wxtom1*wm1toy; mxtom2toy = mxtom2*mm2toy; wxtom2toy = wxtom2*wm2toy; mxtom3toy = mxtom2*mm2toy; wxtom3toy = wxtom3*wm3toy; mxtom4toy = mxtom2*mm2toy; wxtom4toy = wxtom4*wm4toy; RUN; TITLE1; MPLUS Multiple-Group Path Model (all parameters separate by gender; see syntax for full code):! Model code: ON = Y ON X, WITH = correlation! Labels in parentheses (can be used to name constraints between groups) MODEL:! Bring X into the likelihood by estimating its mean and variance [Mind1C] (mxint); Mind1C (mxvar);! Intercepts and residual variances for other variables [Intern Extern Hostile Benev NonTrad] (mm1int mm2int mm3int mm4int myint); Intern Extern Hostile Benev NonTrad (mm1var mm2var mm3var mm4var myvar);! Direct Mind1C --> NonTrad NonTrad ON Mind1C (mxtoy);! Left side of model Intern ON Mind1C (mxtom1); Extern ON Mind1C (mxtom2); Hostile ON Mind1C (mxtom3); Benev ON Mind1C (mxtom4);! Right side of model NonTrad ON Intern (mm1toy); NonTrad ON Extern (mm2toy); NonTrad ON Hostile (mm3toy); NonTrad ON Benev (mm4toy);! Residual covariances among mediator variables Intern Extern Hostile Benev WITH Intern Extern Hostile Benev;! Testing indirect effects MODEL CONSTRAINT: NEW (mxtom1y mxtom2y mxtom3y mxtom4y); mxtom1y = mxtom1*mm1toy; mxtom2y = mxtom2*mm2toy; mxtom3y = mxtom3*mm3toy; mxtom4y = mxtom4*mm4toy;
MODEL Women:! Bring X into the likelihood by estimating its mean and variance [Mind1C] (wxint); Mind1C (wxvar);! Intercepts and residual variances for other variables [Intern Extern Hostile Benev NonTrad] (wm1int wm2int wm3int wm4int wyint); Intern Extern Hostile Benev NonTrad (wm1var wm2var wm3var wm4var wyvar);! Direct Mind1C --> NonTrad NonTrad ON Mind1C (wxtoy);! Left side of model Intern ON Mind1C (wxtom1); Extern ON Mind1C (wxtom2); Hostile ON Mind1C (wxtom3); Benev ON Mind1C (wxtom4);! Right side of model NonTrad ON Intern (wm1toy); NonTrad ON Extern (wm2toy); NonTrad ON Hostile (wm3toy); NonTrad ON Benev (wm4toy);! Residual covariances among mediator variables Intern Extern Hostile Benev WITH Intern Extern Hostile Benev;! Testing indirect effects MODEL CONSTRAINT: NEW (wxtom1y wxtom2y wxtom3y wxtom4y); wxtom1y = wxtom1*wm1toy; wxtom2y = wxtom2*wm2toy; wxtom3y = wxtom3*wm3toy; wxtom4y = wxtom4*wm4toy; SPLH 861 Example 8 page 6 STATA SEM Multiple-Group Path Model (all parameters separate by gender): * first ran using coeflegend, which displays parameter labels (see log) display as result "Multiple group path model omitting coeflegend" display as result "Adding indirect effects (that use parm labels)" sem (mindc -> nontrad) /// XtoY (mindc -> intern extern hostile benev ) /// XtoM1,M2,M3,M4 (intern extern hostile benev -> nontrad), /// M1,M2,M3,M4toY means(mindc) /// print mean var(mindc) /// print variance covariance(e.intern*e.extern e.intern*e.hostile) /// residual covariances covariance(e.intern*e.benev e.extern*e.hostile) /// residual covariances covariance(e.extern*e.benev e.hostile*e.benev) /// residual covariances method(mlmv) /// full-information ML group(sexmw) ginvariant(none), // none= full non-invariance estat eqgof, // R2 per variable estat gof, stats(all), // model fit estat ginvariant, // Wald or Score test for each parm's invariance // Wald = test of constraining equal if unequal // Score = test of allowing unequal if equal // men and women indirect effect XtoM1toY nlcom _b[intern:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.intern] nlcom _b[intern:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.intern] // men and women indirect effect XtoM2toY nlcom _b[extern:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.extern] nlcom _b[extern:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.extern] // men and women indirect effect XtoM3toY nlcom _b[hostile:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.hostile] nlcom _b[hostile:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.hostile] // men and women indirect effect XtoM4toY nlcom _b[benev:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.benev] nlcom _b[benev:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.benev]
STATA SEM unstandardized output for a Multiple-Group Path Model: Structural equation model Number of obs = 653 Grouping variable = sexmw Number of groups = 2 Estimation method = mlmv Log likelihood = -5332.2072 ---------- OIM ---------- Structural nontrad <- intern 0.5480012.1097146 4.99 0.000.3329645.763038 1.4488746.0970253 4.63 0.000.2587086.6390407 extern 0.0473828.1130167 0.42 0.675 -.1741259.2688915 1.0836082.0987572 0.85 0.397 -.1099524.2771687 hostile 0 -.8451654.1564635-5.40 0.000-1.151828 -.5385025 1 -.5347936.1499501-3.57 0.000 -.8286905 -.2408967 benev 0 -.1578276.1593142-0.99 0.322 -.4700778.1544226 1 -.1591458.1440195-1.11 0.269 -.4414189.1231274 0.2131587.3014609 0.71 0.480 -.3776937.8040111 1 -.1263696.239163-0.53 0.597 -.5951205.3423813 _cons 0 6.814568 1.124616 6.06 0.000 4.610361 9.018774 1 7.172311.9379938 7.65 0.000 5.333877 9.010746 -------- intern <- 0.633305.1905479 3.32 0.001.2598381 1.006772 1.0986061.1390832 0.71 0.478 -.173992.3712041 _cons 0 4.39111.1761513 24.93 0.000 4.04586 4.73636 1 5.4137.1332709 40.62 0.000 5.152494 5.674906 -------- extern <- 0.2727332.1725529 1.58 0.114 -.0654644.6109308 1 -.1166402.1378333-0.85 0.397 -.3867886.1535081 _cons 0 3.870788.1594745 24.27 0.000 3.558224 4.183352 1 4.199539.1316457 31.90 0.000 3.941519 4.45756 -------- hostile <- 0 -.1714754.1238283-1.38 0.166 -.4141744.0712236 1 -.1813362.0846223-2.14 0.032 -.3471929 -.0154795 _cons 0 4.333876.1144726 37.86 0.000 4.109514 4.558238 1 3.852628.0810859 47.51 0.000 3.693702 4.011553 -------- benev <- 0 -.3137449.1103874-2.84 0.004 -.5301002 -.0973897 1.1383824.0874718 1.58 0.114 -.0330592.3098239 _cons 0 4.459746.1020472 43.70 0.000 4.259737 4.659755 1 3.848452.0838163 45.92 0.000 3.684175 4.012729 ---------- mean(mindc) 0.8168498.0261972 31.18 0.000.7655043.8681953 1.8471805.0229674 36.89 0.000.8021651.8921958 ---------- SPLH 861 Example 8 page 7
var(e.nontrad) 0 4.124248.356281 3.481866 4.885144 1 4.230226.3073067 3.66883 4.877525 var(e.intern) 0 1.857129.1589556 1.570312 2.196333 1 1.473471.1068968 1.278171 1.698612 var(e.extern) 0 1.522092.1308062 1.286145 1.801324 1 1.433397.1043813 1.242743 1.653301 var(e.hostile) 0.7842846.0671286.6631588.9275339 1.5454591.0395718.4731617.6288033 var(e.benev) 0.6232648.0533465.5270071.7371039 1.5828119.0422816.5055636.6718636 var(mindc) 0.1873576.0160363.1584219.2215784 1.2004512.0145423.1738826.2310795 ---------- cov(e.intern,e.extern) 0.6395836.1101316 5.81 0.000.4237297.8554375 1.5559517.0801768 6.93 0.000.398808.7130954 cov(e.intern,e.hostile) 0 -.4745932.0784874-6.05 0.000 -.6284257 -.3207606 1 -.1836749.0469449-3.91 0.000 -.2756853 -.0916645 cov(e.intern,e.benev) 0.1068751.0654347 1.63 0.102 -.0213747.2351248 1 -.0116527.047542-0.25 0.806 -.1048333.081528 cov(e.extern,e.hostile) 0.0579838.0667428 0.87 0.385 -.0728297.1887972 1.0230488.0454018 0.51 0.612 -.0659371.1120347 cov(e.extern,e.benev) 0.1540697.0598145 2.58 0.010.0368354.271304 1.1566726.0476864 3.29 0.001.063209.2501362 cov(e.hostile,e.benev) 0.035857.0423704 0.85 0.397 -.0471874.1189014 1.1184571.0295551 4.01 0.000.0605301.1763841 ---------- LR test of model vs. saturated: chi2(0) = 0.00, Prob > chi2 =. Group #1 (sexmw=0; N=273) Equation-level goodness of fit Variance depvars fitted predicted residual R-squared mc mc2 observed nontrad 5.828462 1.704215 4.124248.2923952.5407358.2923952 intern 1.932273.0751445 1.857129.0388892.1972033.0388892 extern 1.536028.0139363 1.522092.0090729.095252.0090729 hostile.7897936.005509.7842846.0069753.0835181.0069753 benev.6417075.0184427.6232648.0287401.1695289.0287401 overall.0757179 mc = correlation between depvar and its prediction mc2 = mc^2 is the Bentler-Raykov squared multiple correlation coefficient Group #2 (sexmw=1; N=380) Equation-level goodness of fit Variance depvars fitted predicted residual R-squared mc mc2 observed nontrad 4.853467.623241 4.230226.1284115.3583455.1284115 intern 1.47542.001949 1.473471.001321.0363455.001321 extern 1.436124.0027271 1.433397.0018989.0435769.0018989 SPLH 861 Example 8 page 8
hostile.5520505.0065914.5454591.0119399.1092697.0119399 benev.5866505.0038386.5828119.0065432.0808901.0065432 overall.0283625 mc = correlation between depvar and its prediction mc2 = mc^2 is the Bentler-Raykov squared multiple correlation coefficient ---------------------------------------------------------------------------- Fit statistic Value Description Likelihood ratio chi2_ms(0) 0.000 model vs. saturated p > chi2. chi2_bs(30) 399.257 baseline vs. saturated p > chi2 0.000 Population error RMSEA 0.000 Root mean squared error of approximation 90% CI, lower bound 0.000 upper bound 0.000 Information criteria AIC 10764.414 Akaike's information criterion BIC 10988.493 Bayesian information criterion Baseline comparison CFI 1.000 Comparative fit index TLI 1.000 Tucker-Lewis index Size of residuals CD 0.055 Coefficient of determination ---------------------------------------------------------------------------- Note: pclose is not reported because of multiple groups. Note: SRMR is not reported because of missing values. SPLH 861 Example 8 page 9 Tests for group invariance of parameters ----------- Wald Test Score Test chi2 df p>chi2 chi2 df p>chi2 ----------- Structural nontrad <- intern 0.458 1 0.4985... extern 0.058 1 0.8093... hostile 2.051 1 0.1521... benev 0.000 1 0.9951... 0.779 1 0.3776... _cons 0.060 1 0.8070... --------- intern <- 5.137 1 0.0234... _cons 21.432 1 0.0000... --------- extern <- 3.109 1 0.0779... _cons 2.527 1 0.1119... --------- hostile <- 0.004 1 0.9476... _cons 11.769 1 0.0006... --------- benev <- 10.305 1 0.0013... _cons 21.428 1 0.0000... ----------- var(e.nontrad) 0.051 1 0.8218... var(e.intern) 4.011 1 0.0452... var(e.extern) 0.281 1 0.5961... var(e.hostile) 9.393 1 0.0022... var(e.benev) 0.353 1 0.5523... -----------
SPLH 861 Example 8 page 10 cov(e.intern,e.extern) 0.377 1 0.5393... cov(e.intern,e.hostile) 10.119 1 0.0015... cov(e.intern,e.benev) 2.147 1 0.1428... cov(e.extern,e.hostile) 0.187 1 0.6652... cov(e.extern,e.benev) 0.001 1 0.9729... cov(e.hostile,e.benev) 2.557 1 0.1098... -----------. // men and women indirect effect XtoM1toY. nlcom _b[intern:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.intern] _nl_1: _b[intern:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.intern] _nl_1.3470519.1254253 2.77 0.006.1012229.592881. nlcom _b[intern:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.intern] _nl_1: _b[intern:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.intern] _nl_1.0442618.0631597 0.70 0.483 -.079529.1680526. // men and women indirect effect XtoM2toY. nlcom _b[extern:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.extern] _nl_1: _b[extern:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.extern] _nl_1.0129229.0318837 0.41 0.685 -.0495681.0754138. nlcom _b[extern:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.extern] _nl_1: _b[extern:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.extern] _nl_1 -.0097521.0162089-0.60 0.547 -.0415209.0220168. // men and women indirect effect XtoM3toY. nlcom _b[hostile:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.hostile] _nl_1: _b[hostile:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.hostile] _nl_1.1449251.1080397 1.34 0.180 -.0668289.356679. nlcom _b[hostile:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.hostile] _nl_1: _b[hostile:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.hostile] _nl_1.0969775.0527961 1.84 0.066 -.006501.200456. // men and women indirect effect XtoM4toY. nlcom _b[benev:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.benev] _nl_1: _b[benev:0bn.sexmw#c.mindc]*_b[nontrad:0bn.sexmw#c.benev] _nl_1.0495176.0529333 0.94 0.350 -.0542298.153265
. nlcom _b[benev:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.benev] _nl_1: _b[benev:1.sexmw#c.mindc]*_b[nontrad:1.sexmw#c.benev] _nl_1 -.022023.0243101-0.91 0.365 -.06967.025624 SAS PROC CALIS Multiple-Group Path Model (XtoM1toY path now equal across genders): only new parts of code shown TITLE1 "Multiple-Group Path Model -- XtoM1toY equal by gender"; TITLE2 "Model chi-square provides test of 2 new constraints"; MODEL 1 / GROUP=1; PATH MindC ---> NonTrad = wxtoy, MindC ---> Intern Extern Hostile Benev = XtoM1 mxtom2 mxtom3 mxtom4, Intern Extern Hostile Benev ---> NonTrad = M1toY mm2toy mm3toy mm4toy; MODEL 2 / GROUP=2; PATH MindC ---> NonTrad = wxtoy, MindC ---> Intern Extern Hostile Benev = XtoM1 wxtom2 wxtom3 wxtom4, Intern Extern Hostile Benev ---> NonTrad = M1toY wm2toy wm3toy wm4toy; MPLUS Multiple-Group Path Model (XtoM1toY path now equal across genders): only new parts of code shown MODEL: Intern ON Mind1C (XtoM1); NonTrad ON Intern (M1toY); MODEL Women: Intern ON Mind1C (XtoM1); NonTrad ON Intern (M1toY); SPLH 861 Example 8 page 11 STATA SEM Multiple-Group Path Model (all parameters separate by gender): only new parts of code shown display as result "Multiple group path model omitting coeflegened" display as result "Testing equality of indirect effect XtoM1toY" display as result "Model chi-square gives test of 2 new constraints" sem (mindc -> nontrad) /// XtoY (0: mindc -> intern@a extern hostile benev) /// XtoM1,M2,M3,M4 for group0 (1: mindc -> intern@a extern hostile benev) /// XtoM1,M2,M3,M4 for group1 (0: nontrad <- intern@b extern hostile benev) /// M1,M2,M3,M4toY for group0 (1: nontrad <- intern@b extern hostile benev), /// M1,M2,M3,M4toY for group1 STATA SEM unstandardized output for a Multiple-Group Path Model (relevant parts only): Log likelihood = -5334.9925 ( 1) [nontrad]0bn.sexmw#c.intern - [nontrad]1.sexmw#c.intern = 0 ( 2) [intern]0bn.sexmw#c.mindc - [intern]1.sexmw#c.mindc = 0 ---------- OIM ---------- Structural nontrad <- intern [*].4923674.0727678 6.77 0.000.3497452.6349897 -------- intern <- [*].283537.1135763 2.50 0.013.0609316.5061425 ---------- Note: [*] identifies parameter estimates constrained to be equal across groups. LR test of model vs. saturated: chi2(2) = 5.57, Prob > chi2 = 0.0617